Age | Commit message (Collapse) | Author | |
---|---|---|---|
2022-06-11 | Remove CAN specific LED code | norly | |
This is now handled by common netdev LED events instead: https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next.git/commit/?id=6c1e423a3c84953edcf91ff03ab97829b287184a | |||
2022-06-11 | elm327_init: Update old comment referencing bittiming_const | norly | |
This has been gone since a44b237ce7e9 in 2019. | |||
2022-06-11 | Remove fixed array size for can327_bitrate_const | norly | |
No idea why this was even there. | |||
2022-06-11 | Stop leaking SKBs in elm327_parse_frame() | norly | |
SKBs need to be used or freed before returning, no exceptions. | |||
2022-06-11 | Remove unused includes | norly | |
2022-06-06 | Increase ELM327_SIZE_RXBUF to support fat UART buffers | norly | |
See: https://github.com/norly/elmcan/issues/8 There, a VM was used, and apparently buffers of over 256 bytes were being piped in. | |||
2022-06-06 | On RX buffer overflow, printk remaining input size | norly | |
If can327_ldisc_rx() is called with a large 'count' because the UART driver feeds us huge buffers, then the user should have a chance to know and report this. In this case, we'll have to increase ELM327_SIZE_RXBUF. | |||
2022-06-06 | Fix compilation on Linux up to v5.4 | norly | |
mailbox_read()'s type signature was changed in 4e9c9484b085 which is upstream since v5.5. | |||
2022-05-31 | Accelerate parsing by not re-checking previous data | norly | |
While waiting for a <CR> the code would search through prevois received characters, too. Keep track of what we've already checked to speed this up. | |||
2022-05-31 | Speed up can327_is_valid_rx_char() with a LUT | norly | |
2022-05-31 | Make elm327_rxbuf_cmp() return bool | norly | |
2022-05-31 | Simplify/clarify branching and locking in several places | norly | |
2022-05-19 | Don't ____cacheline_aligned the TX buffer | norly | |
TTY drivers don't seem to care, and if they did, we'd likely have to kmalloc() as before anyway. | |||
2022-05-19 | Style and comments cleanup | norly | |
2022-05-12 | Rename elmcan to can327 | norly | |
This is to clarify that this driver is not a product of ELM Electronics. | |||
2022-05-03 | Restore WAKEUP set_bit vs. write() order in elm327_send() | norly | |
Also, fix a silly spelling error. | |||
2022-05-03 | Repack struct elmcan | norly | |
2022-05-03 | Clear TTY_DO_WRITE_WAKEUP bit on uart_side_failure | norly | |
2022-05-03 | checkpatch.pl style fixes | norly | |
2022-05-03 | Fix lockdep_assert parameters | norly | |
&elm->lock instead of elm->lock | |||
2022-05-01 | Shorten the derivation comment at the start | norly | |
2022-05-01 | Update comment about N_DEVELOPMENT | norly | |
N_DEVELOPMENT has been officially set to 29 in Linux v5.18-rc1. | |||
2022-04-29 | Only netif_wake_queue() when ELM327 is really idle | norly | |
That is, when it is receiving, as that's the default state the driver tries its best to keep it in. | |||
2022-04-28 | Statically allocate TX buffer | norly | |
And also optimise struct elmcan{}'s size with pahole. | |||
2022-04-28 | Start/stop TTY on netdev open/close | norly | |
This allows the ldisc to reset cleanly, and even after a UART error. It also makes for a cleaner ldisc close(). | |||
2022-04-28 | Name enum elm327_to_to_do_bits in lowercase for kernel code style | norly | |
2022-04-28 | Rename hw_failure to uart_side_failure for easier understanding | norly | |
2022-04-28 | elm327_handle_prompt(): Explain size of local buffer | norly | |
Also use snprintf() instead of sprintf()/strcpy(), to clarify the size restriction. | |||
2022-04-28 | Clarify memory/string comparisons | norly | |
We really can't use strncmp() here. Sorry! | |||
2022-04-28 | Remove .hangup() | norly | |
We were just calling .close(), which the TTY layer does for us anyway. And it does so with the correct locking. | |||
2022-04-28 | Simplify TTY sending code and locking | norly | |
The great get_elm() dance was unnecessary, since the TTY layer already synchronises *all* calls to ldisc_ops with tty_ldisc_ref() and tty_ldisc_lock(). All that we have to ensure is that the worker does not race against close(), which is trivial if we clear TTY_DO_WRITE_WAKEUP and then flush the worker. Also, the worker no longer checks whether the netdev is up - all it needs to do is to write out the TTY buffer, and wake up the netdev when the buffer is empty. | |||
2022-03-17 | Get rid of dummy mailbox_read() for rx_offload | norly | |
Use can_rx_offload_add_manual() instead of can_rx_offload_add_fifo(), starting with Linux 5.10. | |||
2022-03-17 | Add can_dropped_invalid_skb() to elmcan_netdev_start_xmit() | norly | |
2022-03-17 | Drop superflouous elm->can.state = CAN_STATE_STOPPED; | norly | |
2022-03-17 | Minor cleanups | norly | |
2022-03-16 | Convert comments to lockdep_assert_held() | norly | |
2022-03-16 | Clean up types and unused vars in struct elmcan | norly | |
2022-03-16 | Rename ELM_ to ELM327_STATE_ | norly | |
2022-03-16 | Unify indentation in struct elmcan | norly | |
2022-03-16 | Move MODULE_* to end of file | norly | |
2022-03-16 | Remove braces from pr_fmt | norly | |
2022-03-16 | Re-add author names | norly | |
2022-03-12 | Rework elm327_hw_failure() | norly | |
Using es58x as a rough template | |||
2022-03-12 | Minor cleanups | norly | |
2022-03-12 | Replace ->can_dlc with ->len | norly | |
can_frame.can_dlc has been deprecated in favour of can_frame.len and as a new driver, this needs to stick to the new convention. Also include the ugliest backwards hack #define I've ever written to make this compile on Linux <= 5.10 which is currently LTS. | |||
2022-03-12 | elmcan_netdev_start_xmit(): Remove unnecessary skb->len check | norly | |
linux/net/can/af_can.c's can_send() guarantees this already. | |||
2022-03-12 | Use alloc_can_skb() | norly | |
2022-03-12 | Rename TODO_* to ELM327_TX_DO_* | norly | |
Also rename elm->can_frame to elm->can_frame_to_send | |||
2022-03-07 | Mark {_len,}_memstrcmp inline | norly | |
2022-03-07 | Change MAGIC to DUMMY, and silence init | norly | |